linux-kernel - Re: [PATCH 3/4] dt-bindings: nvmem: Allow bit offsets greater than a byte

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 8 Sep 2022 22:29:13 -0500
From:   Samuel Holland <samuel@...lland.org>
To:     Rob Herring <robh@...nel.org>
Cc:     Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Chen-Yu Tsai <wens@...e.org>,
        Jernej Skrabec <jernej.skrabec@...il.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-kernel@...r.kernel.org, linux-sunxi@...ts.linux.dev
Subject: Re: [PATCH 3/4] dt-bindings: nvmem: Allow bit offsets greater than a
 byte

On 8/25/22 4:02 PM, Rob Herring wrote:
> On Sun, Aug 14, 2022 at 12:36:54PM -0500, Samuel Holland wrote:
>> Some NVMEM devices contain cells which do not start at a multiple of the
>> device's stride. However, the "reg" property of a cell must be aligned
>> to its provider device's stride.
> 
> How is a DT author supposed to know this? 

To know the stride? They know the compatible string of the NVMEM provider
device, and the stride is a property of the hardware.

To know that alignment to the stride is required? That's not documented in the
binding. I can add that to the "reg" description in this file.

> I would lean toward it's the OS's problem to deal with (your option 1 I 
> guess). I worry that one client could expect it one way and another 
> client the other. Or folks making DT changes to 'fix' things.

After this binding change, the meaning of

	reg = <0x2a 1>;
	bits = <0 8>; // optional

and

	reg = <0x28 4>;
	bits = <16 8>;

would be identical.

With option 1, only the first representation would be valid in the DT.

With this series (option 2), either representation would validate in the DT, but
only the second representation would be accepted by Linux for the driver in
question (sunxi-sid).

With option 3, either representation would work in the DT and Linux.

Note that there is no restriction on the bit length, so the following are
already equivalent today:

	reg = <0x28 1>;
	bits = <0 8>; // optional

and

	reg = <0x28 2>;
	bits = <0 8>;

So there are already multiple possible representations. I don't really think
that is a problem, since the meaning is still unambiguous.

Regards,
Samuel

>> These cells can be represented in the DT using the "bits" property if
>> that property allows offsets up to the full stride. 63 is chosen
>> assuming that NVMEM devices will not have strides larger than 8 bytes.
>>
>> Signed-off-by: Samuel Holland <samuel@...lland.org>
>> ---
>>
>>  Documentation/devicetree/bindings/nvmem/nvmem.yaml | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/Documentation/devicetree/bindings/nvmem/nvmem.yaml b/Documentation/devicetree/bindings/nvmem/nvmem.yaml
>> index 3bb349c634cb..4f440ab6a13c 100644
>> --- a/Documentation/devicetree/bindings/nvmem/nvmem.yaml
>> +++ b/Documentation/devicetree/bindings/nvmem/nvmem.yaml
>> @@ -53,7 +53,7 @@ patternProperties:
>>          $ref: /schemas/types.yaml#/definitions/uint32-array
>>          items:
>>            - minimum: 0
>> -            maximum: 7
>> +            maximum: 63
>>              description:
>>                Offset in bit within the address range specified by reg.
>>            - minimum: 1
>> -- 
>> 2.35.1
>>
>>